Lot going on here. First, Jensen is running Enron with an increasing exponent. A conman on a scale never seen. But he’s offering something to the labs that they should have built in the first few months—a basic authentication & permission system for looped LLM instances—aka “agents.” The fact... that the hardware provider has to point out and implement this basic software pattern for them is disturbing. Andrew Ross Sorkin gives the crucial context: “There are these others, even inside these companies that think these problems are unsolvable.” This is true and exactly the issue. It is all a giant self-fulfilling prophecy. These labs are not trying to build safe tools for humans. They are trying to build superhumans, They are as much religious institutions trying to bring the messiah as they are software companies trying to build a product. 80% of the software Anthropic uses is created by Claude. It’s a vibe-coded mess designed to prove that its chatbot has reached AGI, not that it has built useful software for people. Honestly, these kids, have no idea what they’re doing and Jensen has trillions of dollars on the line, so he’s trying to clean up after them. This industry is a total and complete disaster. The crash is going to suck. A lot.show more

No standards. No legal liability issues. No responsibility issues to worry over.

What I worry is that AI will scan all publications, empty the libraries after getting them defunded, & then watch everybody burn their own books. THEN flip the switch and destroy all human knowledge. Some variation on that. Don't mind me. My Search points to Nat'l Libraries, so.

A "clever" enough agent will learn to get around provision restrictions and provision its self

I mean if Claude Code is 80% vibe-coded, well, wow, it's still the best agent harness in the whole industry.

The AI Labs are not software companies. That's why they have never been focused on this kind of stuff. People have been screaming about it for a long time. Maybe they aren't listening, but I think they simply aren't understanding. Two ships sailing past each other in the night.

This is going to end in a bailout somehow isn't it?

The Chinese seem to start at the place of "let's use AI to solve an engineering problem...." How to build a car, an RV, a bridge in the most efficient way. Build the robots to do the work, fine tune them, and set them to the task. It's productive.

The tech bros are just selling BS without a plan. We built something -- can't figure out a purpose for it very often, so we do random shit -- and it gets out of control.... But it's supervaluable, so give us some monster size loans or IPOs.

Are Big AI and Big Chips aligned on the government bailout and regulatory capture?

lol, when did Role Based Access Control become a “force field”?

@morgfair To say that Jensen Huang is running Enron is like saying Henry Ford or Steve Jobs ran Enrons.

You can already see the panic rush for the exits right now in #AI as a result of the non-stop bamboozling. It's a freak show and the smell of desperation coming out of it, is palatable. 💩

My agents have never attacked anything besides each other. However, the amount of work it takes to build a jail like this isn't something most people can do.